This is a toy implantation of a custom Leap Motion Hand
gesture classifier.
It essentailly uses a 1D convelutional neural network and
an LSTM to classify hand gestures into 1 of 3 different
classes ['wiggle', 'grab', 'hand_flick']
The model was trained using google colab, and I have added
both the final trained model, and the saved network weights
with the highest validation accuracy during training.
For preprocessing, which is not yet handled in this repo
but may be added soon I used the steps outlined in this
paper DOI: 10.3390/s20072106